Lecture 9 Registration with ITK

This work by John Galeotti and Damion Shelton, © 2004-2013, was made possible in part by NIH NLM contract# HHSN276201000580P, and is licensed under a Creative Commons Attribution 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by/3.0/ or send a letter to Creative Commons, 171 2nd Street, Suite 300, San Francisco, California, 94105, USA. Permissions beyond the scope of this license may be available by emailing [email protected] most recent version of these slides may be accessed online via http://itk.galeotti.net/

Lecture 9Registration with ITK

Methods in Medical Image Analysis - Spring 2013 BioE 2630 (Pitt) : 16-725 (CMU RI)

18-791 (CMU ECE) : 42-735 (CMU BME)

By Dr. John Galeotti & Dr. Damion Shelton

1

http://creativecommons.org/licenses/by/3.0/

2

For more info/gory detail…

Please see the following for exhaustive detail: Chapter 8 in the ITK Software Guide

Currently, this does not cover the newer v4 registration methods Insight into Images ITK Source Tree

Examples/Registration/ E.g. Examples/Registration/ImageRegistration1.cxx

ITK Doxygen http://www.itk.org/Doxygen43/html/

group__RegistrationFilters.html http://www.itk.org/Doxygen43/html/group__Group-

Registration.html http://www.itk.org/Doxygen43/html/group__Group-Numerics.html

3

What is registration?

The process of aligning a target image to a source image

More generally, determining the transform that maps points in the target image to points in the source image

4

Transform types

Rigid (rotate, translate)Affine (rigid + scale & shear)Deformable (affine + vector field)Many others

5

Registration in ITK

ITK uses an extensible registration frameworkVarious interchangeable classes existRelatively easy to “twiddle” the part you’re interested

in while recycling prior work

The new ITKv4 Registration components use a separate (but similar) framework (discussed later)

SimpleITK currently does NOT support registration. For “simple” usage, consider using ANTS instead: http://www.picsl.upenn.edu/ANTS/

6

ITK terminology

Fixed image f(x) - stationary in spaceMoving image m(x) - the fixed image with an

unknown transform appliedGoal: recover the transform T(x) which maps

points in f(x) to m(x)

7

Registration framework pieces

2 input images, fixed and movingMetric - determines the “fitness” of the current

registration iterationOptimizer - adjusts the transform in an attempt

to improve the metricInterpolator - applies transform to image and

computes sub-pixel values

8

Figure 8.2 from the ITK Software Guide v 2.4, by Luis Ibáñez, et al.

ITK registration flowchart

9

ITK’s “Hello world” example

2D floating point image inputsPlease see the software guide (section 8.2) for

code specificsI am going to cover what each piece does, not

look at code per se

10


ITK’s “Hello World” Example:Flow Chart for Everything

11

Input images

2D floating pointFloating point avoids loss of precision problems

with integer pixel types

12

Transform

TranslationTransformPermits translation only in 2DDocumentation notes similarity to

AffineTransform, but improved speed (why…?)

13


Transform Coordinate System

14

Metric

MeanSquaresImageToImageMetricSum of squared differences between 2 images

on a pixel-by-pixel basisA bit naïveWorks for 2 images that were acquired with the

same imaging modality

15

Optimizer

RegularStepGradientDescentFollows the derivative of the metricStep size depends on rapid changes in the

gradient’s directionStep size eventually reaches a user-defined

value that determines convergence

16

Interpolator

LinearInterpolateImageFunctionFast and conceptually simple

17

Wrapper

ImageRegistrationMethodCombines all of the previous classes into a

master class

registration->SetMetric( metric );registration->SetOptimizer( optimizer );registration->SetTransform( transform );registration->SetInterpolator( interpolator);

18

Other steps

Set the region of the fixed image the registration will operate on (e.g. to ignore bad data)

Initialize the transformTwiddle the optimizer for best performance*

*may involve pain and suffering

19


Hello world input

20

Figure 8.6 (top) from the ITK Software Guide v 2.4, by Luis Ibáñez, et al.

X & Y translation vs. time

21

Figure 8.6 (bottom) from the ITK Software Guide v 2.4, by Luis Ibáñez, et al.

Metric vs. time

22

Registration results

After registration converges/terminates you call GetLastTransformParamters to recover the final transform

For the Hello World example there are 2 parameters, X & Y translation

23

Double checking results

Use ResampleImageFilter to apply the transform for the fixed image

Take the output, compute a difference image with the moving image, examine the results

Good registration results in nothing much to see

24


Image comparison

Difference beforeregistration

Difference afterregistration

Registeredmoving image

25

Keeping tabs on registration

Registration is often time consumingIt’s nice to know that your algorithm isn’t just

spinning it’s wheelsUse the observer mechanism in ITK to monitor

progress See the software guide, 3.2.6 and 8.4

We’ll see this again later, when we discuss how to write your own ITK filters itk::ProgressEvent is one example

26

Observer steps

Write an observer class that will process “iteration” events (Just copy some code from an example)

Add the observer to the optimizerAs a generic note, observers can observe any class

derived from itk::ObjectStart registration as usual

27

Things observers can do

Print debugging infoUpdate GUIOther small management functionsShould not do anything too processor intensive

28

Multi-modality registration

Remember how I said sum-of-squares difference is relatively naïve?

Mutual information helps overcome this problem

Section 8.5 shows how to implement a simple MI registration

29

Notes about the MI example

Significantly, largely the same piece of code as Hello World

Mutual Information is a metric, so we can keep the optimizer, the interpolator, and so on

Majority of differences are in tweaking the metric, not in rewriting code

30


MI Inputs

T1 MRI Proton density MRI

31Taken from Figure 8.10 of the ITK Software Guide v 2.4, by Luis Ibáñez, et al.

MI Output: Image Comparison

Before AfterThis is an example of a checkerboard visualization

32

Centered transforms

More natural (arguably) reference frame than having the origin at the corner of the image

Details are not appreciably different from other rigid registrations, see 8.6

33

An aside: “Twiddling”

A common criticism of many/most registration techniques is their number of parameters

A successful registration often depends on a very specific fine-tuning of the algorithm

“Generalized” registration is an open problem

34

Multi-Resolution registration

Useful to think of this as algorithmic “squinting” by using image pyramids

Start with something simple and low-resUse low-res registration to seed the next higher

stepEventually run registration at high-resAlso called “coarse to fine”

35


Multi-resolution schematic

36


Image pyramids

37

Optimization

Parameter dependency rears its ugly headYou often/usually need to adjust optimizer

parameters as you move through the pyramidYou can do this using the Observer mechanism

38

Multi-resolution example

Again, mostly the same code as Hello WorldUse MultiResolutionPyramidImage filter to

build fixed and moving pyramidsMultiResolutionImageRegistrionMethod is now

the overall framework

39

Benefits of multi-resolution

Often fasterMore tolerant of noise (from “squinting”)Minimizes initialization problems to a certain

extent, though not perfect

40

See the software guide for…

Detailed list of: Transforms Optimizers Interpolation methods

You’re encouraged to mix and match!

Note: ITKv4’s new registration methods are in separate classes, and most available documentation does not cover them yet.

Check Doxygen and ITK source code for, e.g.: ImageRegistrationMethodv4 MattesMutualInformationImageToImageMetricv4 DemonsImageToImageMetricv4 QuasiNewtonOptimizerv4 (an improved gradient descent)

41

Deformable registration

ITK has 2 primary techniques: Finite element: treat small image regions as having

physical properties that control deformationDemons: images are assumed to have iso-intensity

contours (isophotes); image deformations occur by pushing on these contours

42

Model based registration

Build a simplified geometric model from a training set

Identify parameters that control the characteristics of the model

Register the model to a target image to adapt to a particular patient

43

Model based, cont.

Uses the Spatial Objects framework for representing geometry

Useful because it derives analytical data from the registration process, not just a pixel-to-pixel mapping

44Figure 8.60 from the ITK Software Guide v 2.4, by Luis Ibáñez, et al.

Model-based example

Note: This is what we want, NOT the output of an actual registration

45


Model-based reg. schematic

46

Model-based registration: Warning!

ITK does not yet directly support generic model-based registration “out of the box”

ITKv4 does support point-set to image registrationOtherwise, model-based reg. requires writing your

own custom ITK transform, with new parameters Transform’s new parameters Spatial Object parameters You must individually map your custom transform’s new

parameters to the specific spatial object parameters you want to allow registration to adjust

This isn’t too complicated if you know what you’re doing Search Insight Journal for examples

47

Speed issues

Execution time can vary wildlyOptimizer (more naïve = faster) Image dimensionality (fewer = faster)Transform (fewer DOF = faster) Interpolator (less precise = faster)

48

New in ITKv4 (ImageRegistrationMethodv4, etc.) New unified, improved, and fully multi-threaded optimization and registration

framework (including multi-threaded metrics) Dense deformation fields (including a new transform that encapsulates a dense

deformation field) Point Set registration methods (landmark or label guided registration)

Automatic parameter scale estimation for transforms Automatic step-size selection for gradient-based registration optimizers Composite Transforms (grouping multiple transforms into a single one) Symmetric registration (where the Fixed and Moving images make unbiased

contributions to the registration)

New metrics for Demons and Mutual Information Diffeomorphic (velocity field) deformable registration Additional evolutionary optimizers Improved B-Spline registration approach available and bug fixes to old framework Accurately transform and reorient covariant tensors and vectors

List taken from http://www.itk.org/Wiki/ITK_Release_4/Why_Switch_to_ITKv4 andhttp://www.itk.org/Wiki/ITK_Release_4/Migration_Plan/Release_Notes/ITKv4_Final_Release_Notes

49

ITKv4 Registration

Uses a different framework than “traditional” ITK registration. The new framework is designated with a “v4” suffix.

You must use a v4 metric and a v4 optimizer when doing a v4 registration!

Take a look here: http://www.itk.org/Doxygen43/html/group__ITKRegistrationMethodsv4.html http://www.itk.org/Doxygen43/html/group__ITKMetricsv4.html http://www.itk.org/Doxygen43/html/group__ITKOptimizersv4.html ITK source code: Modules/Registration/RegistrationMethodsv4/include/ ITK source code: Modules/Registration/Metricsv4/include/ ITK source code: Modules/Numerics/Optimizersv4/include/

50

ITKv4 Registration Example Code

ITKv4’s testing files provide some examples of v4 Registration. In particular, look at: Modules/Registration/RegistrationMethodsv4/test/

itkSimpleImageRegistrationTest.cxxThis is a big example. First it creates an observer to print out the ongoing status of the

registration.Then it uses GradientDescentOptimizerv4 to guide an affine

registration. Finally it uses GradientDescentOptimizerv4 again to guide a 2nd-

stage vector-field deformable registration.

Look at some of the other v4 test files as well!

51

Take home messages

Exactly what parameters do what is not always obvious, even if you are familiar with the code

Successful registrations can be something of an art form

Multi-resolution techniques can helpWork within the framework!

Documents

Lecture 9 Registration with ITK